For solving the difficult problems that relate to the unlawful pouring of building rubbish and fragments close to electric power construction places—including dispersed objects, broken dividing lines, big mixing with open building places and lawful treating areas, and slow reacting speeds in traditional inspections—this article puts forward one recognition and early warning system which is based on remote sensing. This system has built a multi-source data arrangement frame which includes an engineering restriction layer, a time remote sensing layer, and a high-resolution detailed check layer. At first, it carries out the definition of the candidate search region on the basis of power line corridors, tower foundations, substation expansion areas, construction access roads, and already approved disposal places. It then uses constraint-guided spatiotemporal boundary refinement networks to perform anomaly screening, three-branch feature coding, and target boundary restoration.Finally, a hierarchical early warning mechanism is established by integrating criteria such as persistence, expansion, deviation from legality, and environmental sensitivity. Results show that CSBR-Net achieves a peak F1 score of 89.2% on the overall test set, with an average F1 of 89.6% across scenarios, and maintains an 87.4% recognition rate even under wet surface conditions; the latency of the complete model is 43.0 ms per image, representing improvements of 4.6 and 7.3 percentage points over the optical baseline in F1 and mIoU, respectively. Typical examples have shown that risk numerical values can pass the alarm and attention critical points before the goal region obviously enlarges, hence giving early notification for on-the-spot checking. This research therefore points out that putting engineering law limit requirements, time change data, and fine divided boundary description into one single decision-making chain can effectively promote the stability and deployable ability of illegal soil and waste dumping inspection and early warning systems that lie near electric power construction projects.
